THE EFFECTS OF PREWSE COM’ENT ON‘ ACCURACY Am SQLUTIGN HMS 3N SYLLOGISHC REASOMNG Thesis. for the fiegrw ef M. A. mama}; STATE UNN’ERSiT‘f GEflRGE L MRROTT 1.965;" THESIS L I B R A R Y Michigan State Unix/usity ABSTRACT THE EFFECTS OF PREMISE CONTENT ON ACCURACY AND SOLUTION TIME IN SYLLOGISTIC REASONING By George L. Parrott This investigation was concerned with the effects of true premise content, false premise content, and mixed premise content on logical reasoning scores for both accuracy and solution time. Three test forms were developed with items of differing truth-falsity content, but matched for logical structure. In contrast to the development of earlier scales, the truth-falsity of each premise was scaled by a sample of subjects, rather than only by the experimenter. Experiment I compared the three test forms for number correct with a sample of 99 undergraduate students randomly assigned to test forms. In support of the hypotheses, the false items were found to be easiest, then the true items, and then the mixed premises items. Experiment II compared the three types of item content for total solution time with lO-item scales matched for word length. A total of 24 subjects were individually tested in this experiment with the items presented with the Johnson Serial Exposure Box. In George L. Parrott terms of total solution time, false items take the longest, then true items, and then mixed premise items. A sub-comparison made in Experiment II concerned the relative transfer effects of the true-then-false presentation order as compared to the false-then-true order. For number correct, the transfer for both orders was identical, but the false-then-true order significantly reduced the solution time for the true items, while the true items did not significantly reduce the solution time of the subsequent false items. The discussion of the results emphasized the suggestions of Henle (1962) in terms of errors in logical reasoning; and Berlyne's model of conceptual conflict best fit the data for both solution time and number correct. Even the Berlyne model was not completely satisfactory, and an attempt was made at intergrating an information processing model with the Henle (1962) and Berlyne (1957) conceptions to derive a theoretical system that handled all three types of item content for both number correct and solution time. Of some methodological and theoretical interest, it was suggested that solution time and number correct are not the interchangeable dependent variables that they have often been assumed to be, but rather for some problems, at least, they are positively correlated measurements. Suggestions for further research investigating differential effects of training procedures and instructions on reasoning with varied item content, and George L. Parrott for investigating the test-retest reliability of solution time scores were made. Approved A;/f.~,v_ ,, I; //’l’ " . ; /'. .' L.“ Chairman, Thesis Committee ) , ,_../ 'i .' ‘ it. I ,. r/ , 1/ Date ,,41 ;LL ; THE EFFECTS OF PREMISE CONTENT ON ACCURACY AND SOLUTION TIME IN SYLLOGISTIC REASONING By George L. Parrott A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1967 To My Wonderful Wife ACKNOWLEDGMENTS The progress and completion of this research owes much to others than the experimenter. Dr. Donald M. Johnson, Committee Chairman, has given guidance, encouragement, and time freely and most beneficially. Dr. James L. Phillips has aided in the design and statistical analyses, and Dr. Milton Rokeach has added insight in theoretical considerations. To each of these men go sincere appreciation. iii TABLE OF CONTENTS Chapter Page I. INTRODUCTION ............. . ............... 1 II. REASONING WITH TRUE AND FALSE PREMISES 8 III. METHOD ................................... 12 Subjects ............................ 12 Material ............................ 12 Apparatus ........................... 14 Procedure ........................... 14 IV. RESULTS .................................. 19 Effect of Premise Content on Number Correct ...................... 19 Effect of Premise Content on Solution Time ....................... 24 V. DISCUSSION ............................... 27 VI. SUMMARY .................................. 34 VII. REFERENCES ............................... 36 iv Table 10. ll. 12. 13. 14. LIST OF TABLES Page Mean length and SD of true, false, and mixed reasoning scales in Exp. II .............. 13 Mean number correct by premise content, Exp. 1, 27—items ............................... l9 ANOV for total score on 27-item tests .......... 19 Mean number correct by premise content, Exp. 1, lS-items ............................... 20 ANOV for total score on lS-item tests .......... 21 Number correct, Experiment II, lO-item tests .......................................... 21 ANOV for total correct on Exp. II, lO-item tests .......................................... 22 Errors separated by item type and premise type ........................................... 23 Chi-square analysis for error distribution ..... 23 Number of errors on T and F scales by position ....................................... 23 Summary of solution time in seconds for Exp. II ........................................ 24 ANOV for solution time in seconds on three test forms ..................................... 24 Effects of presentation order on mean solution time for T and F premise material ..... 2S ANOV summary for comparison of position effects on T and F premise syllogistic reasoning times ................................ 25 LIST OF APPENDICES Appendix Page A. PREMISES SCALING FORMS--3 and 4 .............. A-l B. SYLLOGISTIC REASONING TESTS—-EXP. I .......... B-l C. TABLE OF ITEM NUMBERS FROM EXPERIMENT I FOR MATCHED ITEMS AND ITEMS USED IN EXP. II ...... C-l D. SUBJECT RECORD SHEET USED IN EXP. II ......... D-l vi INTRODUCTION While the logician has asserted ”there are logical principles of psychology (as of every science), but there are no psychological principles of logic” (Windelband, 1961, p. 9), the psychologist must approach the application of logical thinking as an area relevent to his study. Syllogistic reasoning is one special form of logical deduction which has been studied by the psychol- ogist in his research on human cognitive processes. This paper will concentrate on the effects of premise content on syllogistic reasoning. One of the earliest and most comprehensive investigations into variables affecting syllogistic reasoning was the Wilkins (1928) study. This author deve10ped four different twenty-item tests (with alternate forms of each test); these four tests were approximately equated for logical structure. Logical structure refers to the four types of premises commonly called universal affirmative, particular affirmative, universal negative and particular negative. The basic differences between the four tests Wilkins studied were as follows: Test 1 premises dealt with familiar material; Test 2 premises dealt with symbolic material, e.g. all A is B; Test 3 premises dealt with unfamiliar and made up terms, e.g. all zygatropes are metathanes; and Test 4 premises dealt with what Wilkins called "suggestive” material, e.g. all metal is wet. Form 1 and Form 4 are of most interest to us, as these compare what might be called ”true” or factual premises (input) with what might be called ”false” or unreasonable premises. In the suggestive items of the Wilkins study, subjects were required to perform reasoning operations with unreasonable premises. A sample problem from the suggestive items is as follows: All Anglosaxons are English; all British are Anglosaxons; therefore A. All British are English B. All English are British C. Some British are not English Subjects were instructed to mark the correct conclusions with a (+) and the incorrect conclusions with a (-). Wilkins compared percent correct scores on these suggestive items with similar scores on the familiar items and concluded that familiar items were easier. Two critical difficulties in this study are apparent to the careful reader: 1. Wilkins judged suggestive premises herself, and what might be ”suggestive” to an advanced graduate student need not be equally so to her undergraduate subjects; and 2. her Form 1 and Form 4 scales were not exactly matched for logical structure, thus the effects of item content are confounded by unbalanced scales (Thistlewaite, 1950, has shown that ”psychological” structure is an important variable in reasoning distortions). In addition to these design problems, little statistical analysis of the differences obtained between treatment groups was done in this early study. Thurstone (1938) in his factor analysis of primary mental abilities used three types of syllogistic reasoning tests; his false premises test was composed of items similar to the following example: All haystacks are catfish. All catfish are typewriters. Therefore all haystacks are catfish. Q+) These items were judged by the student subjects either (+) for good reasoning or (-) for bad reasoning. Thurstone's second syllogistic reasoning scale was the "reasoning test,” and it was composed of items similar to the one given below. All sports are dangerous, and football is a sport. Therefore football is dangerous. (+) The final deductive reasoning battery of syllogism items was the "syllogism test," and this was designed to investigate reasoning with extremely monotonous material. All of the items in this scale involved Smith, Jones, and Brown. The false premises and the reasoning test correlated .78 with each other in the Thurstone investigation, but the monotonous syllogism test correlated only .42 to .57 with these tests. From the complete factor analysis, a factor called ”verbal relations” was extracted, including the false premises and the reasoning test, but not the monotonous material. Janis and Frick (1943) found that graduate student subjects made more errors in judgments of the logical validity of syllogistic items with acceptable (agreed with) conclusions and invalid arguments than with unacceptable (not agreed with) conclusions and invalid arguments. In addition, when subjects disagreed with the conclusions, but valid arguments were given, more errors were made than in the ”disagree-invalid" condition. These authors did not consider the premise content of their items as Thurstone (1938), and Wilkins (1928) had, rather only the conclusions were of concern. Lefford (1946) developed two matched syllogism scales, one of "emotional” item content, and one of ”nonemotional” content and compared validity judgments (among other things) for these two types of item content. He concluded that neutrally toned items were solved more correctly than emotionally toned items and that solving emotionally toned items first had a negative effect on later solution of neutral items. When neutral items proceeded emotional items, some subjects seemed to exhibit positive transfer on the later emotional items, but here Lefford is not too clear as this conclusion is based on score distributions and not on group means or other more accepted measures of central tendency. No statistical test and no group means are reported for these experiments, and thus these results must be open to review. Morgan and Morton (1943, 1944) have reported several investigations of distortions in logical reasoning as a function of personal convictions. Similar to the work of Janis and Frick (1943), these authors find that the amount and direction of distortion in logical reasoning can often be related to personal attitudes. Thistlewaite (1950) has used this general "distortion" design with items of neutral and ethnocentric content, and in a well designed study confirmed that personal convictions clearly influence sensitivity to distortions in logical thinking. Henle and Michael (1956) have summarized much of this previous "personal conviction influence” literature in their discussion and research on the influence of attitudes on syllogistic reasoning. Using a multiple-choice test format, Henle and Michael replicated the Morgan and Morton (1943, 1944) findings, but they then measured the subjects' attitudes rather than inferring that attitude was the critical factor. Their results agreed with the Morgan and Morton findings, but their attitude measures indicated that attitude was not the critical factor in reasoning distortions. In fact, attitude loaded items were solved correctly slightly more often than symbolic items, thus Henle and Michael concluded that many subjects did not understand the task correctly. In a second experiment, subjects were given quite complete instructions in doing syllogistic reasoning, and significant gains in accuracy of reasoning for both types of material were found. Henle and Michael suggest that research needs to be designed to consider not whether attitudes influence reasoning, but how they influence the reasoning process. Asch (1952) has discussed the interaction of attitude with fact and suggests that "what is needed is observation of reactions to facts that are compatible with one's established view and to facts that contradict it.” Henle (1962) has concluded that there are four basic sources of errors in syllogistic reasoning: 1. failure to accept the logical task; 2. restatement of the premises; 3. omission of a premise; and 4. addition of a premise. Thus Henle sees errors not as a function of logical faults, but rather due to the lack of acceptance of the task as designed. If this interpretation of errors is correct, it implies that more errors would be made with material that could be easily misread or misunderstood by the subjects. This hypothesis could be tested with appropriately designed premise material; for example, items with one true or familiar premise and one false or unfamiliar premise should be more susceptible to misreading or premise restatement than either both true premises or both false premises material. Wilson (1965) has investigated the effects of "competition" instructions, defined in terms of public reporting of individual subjects scores, on syllogistic reasoning, and some of his findings are of direct interest here. Wilson used items from the Ruch, Warren, and Gray (1963) workbook, and his results indicated that ”biased" items took longer to solve and were solved less accurately than familiar items. These results must be held somewhat questionable due to the incomparability of many of the items for logical structure. It should also be noted that subjects simply indicated correct or incorrect for each item, and thus the actual reasoning task was very directly related to the single conclusion presented. Another, if not more reasonable, procedure would have been to have subjects select the correct conclusion from a number of alternatives. Frase (1966), in the most recent published investi- gation in the area of syllogistic reasoning and belief variables, found that error and solution time scores were systematically influenced by varying conclusion statements at three levels of semantic differential incongruity. This summary of the available research in the area of syllogistic reasoning with varied material points out some clear gaps in the experimental literature. While we have some evidence that familiar material is easier than "suggestive" material (Wilkins, 1928; Wilson, 1965) and nonemotional is easier than emotional (Lefford, 1946), neither of these conclusions is based on tests which have been carefully balanced for logical structure, or in case of the Wilson (1965) material, for item length. The work of Henle (1962) and Henle and Michael (1956) has shown that attitudes may influence reasoning, as some of these earlier experimenters concluded, but the unanswered question is how the reasoning process is affected by varying item content. McGuire (1960) has shown that subjects' reasoning processes correspond to a logical model based on the structure of the syllogism. In his research, he had subjects rate statements on a truth-falsity scale from 0 (completely false) to 100 (completely true); these statements were both premises and conclusions to syllogistic arguments, and when subjects received information designed to change the truth-falsity rating of a conclusion, the appropriate changes in the rating of the premises leading to that conclusion also took place. Johnson (in press) has covered much of this previous research in his paper on reasoning and logic, and he concludes that while a logical model of human reasoning is not perfect, no other model works so well. In attempting to analyze the effects of bias on logical reasoning, we Should re-examine the work of Wilkins (1928) and Lefford (1946). As Wilkins admitted (p. 15), some of her suggestive items were not interpreted by her subjects as she had designed them to be interpreted; and Lefford (1946) also deve10ped his emotional items based on his a priori judgment of their emotionality. On close inspection, it might appear that his emotional items could also be considered untrue or half-truths. In reviewing these items, this experimenter had subjects scale the truth-falsity of each premise from the Lefford tests on a scale from one to seven. The nonemotional items had a mean premise value of 2.31 (SD=1.77) and the emotional items had a mean premise value of 3.97 (SD=2.01); this difference in scaled truth-falsity of the Lefford premises was statistically significant (t=3.37, p.<01); and it suggests that the emotionality factor which Lefford attempted to study was confounded by varying truth-falsity of his emotional and non- emotional items. REASONING WITH TRUE AND FALSE PREMISES Reasoning with premises of varying truth-falsity, it has been asserted, occurred in the Lefford (1946) study, the Wilkins (1928) study, and the similar Wilson (1965) research. Reasoning with this type of premise material could be looked at from at least two theoretical positions: 1. the belief-disbelief theory of Rokeach (1960); and 2. the later information processing approach, typified by the work of Suedfeld and Hagen (1966) and Schroeder, Driver and Streufert (1967). Both of these theoretical positions lend them- selves to treatment of the question of reasoning with material of different types. Rokeach (1960) presents a model of cognitive structure called the belief-disbelief system, and he asserts that the belief system is a relatively well interconnected and organized structure as compared with the disbelief system. Rokeach (1960, p. 33) defines the belief system as representing "all the beliefs, sets, expectancies, or hypotheses...that a person...accepts as ” The disbelief system ”is composed of a series true.... of subsystems rather than merely a single one, and contains all the disbeliefs...a person at a given time rejects as false.” While Rokeach's major concern is with individual differences in these belief-disbelief systems, for the cognitive theorist they present a model which might be tested in some general way. False premise syllogisms could be considered disbelief system statements and true premise syllogisms belief system statements; from the general Rokeach model, the disbelief statements would be harder to reason with (take more time) than the belief system statements, but no clear prediction as to errors in reasoning can be made from this theoretical model. The more recent information processing theorists consider information complexity to be the critical environmental variable (Suedfeld and Hagen, 1966); this complexity is determined by: 1. information load; 2. information diversity; and 3. rate of change of information. Many of the workers in the area of information processing approaches to cognition have discussed categories or codes used by the subject in integrating new information or perception (Bruner, 1957a; 1957b; Wallach, 1958). If the true premise material can be considered analogous to available, well used categories, while the false premise material represents poorly used or unavailable categories, it is not to difficult to consider false premises as a relatively high information input load, and a relatively diverse input load as compared to reasoning with true (familiar) statements. Longer total solution time should be required for reasoning with false premises as compared to true premises; again, this model offers no insight for reasoning accuracy. Berlyne (1957) suggests that ”conceptual conflict” may be produced as a function of the conceptual incongruity of the false material (that is, a conflict between learned 10 symbolic response), and that this conflict produces a drive reduction process leading to behavior (search for knowledge). The behavior or search for knowledge would take the form of the active relating of the unfamiliar (false) premises into a coherent argument. Consequently, we might expect more time to be involved in solving false items than true items, but here the implication exists that the extra time should lead to more correct solutions to the false items. In an exploratory study by this experimenter using syllogisms with true and false premises, some evidence was obtained that false premises were encoded (or read) more slowly than true premises, and some indication was found that false items were gotten correct at least as often as true items. This current study will follow up these findings and attempt to further compare reasoning with true and false premises in a larger sample of subjects and to further obtain data for subjects reasoning with syllogistic arguements composed of one true and one false premise. Thus three test forms will be developed, identical in logical form and word lenth, but differing in specific item content: 1. all true premises (True or T); 2. all false premises (False of F); and 3. one true and one false premise in each item (Mixed or M). From the exploratory study, and the previous work in this area, the following hypotheses can be generated and tested: For Number Correct H1. Mixed items will lead to more errors (Henle, 1962). H2. False items will have fewer errors than true items (from exploratory study). 11 For Solution Time H (Note: 3. False items will take longest; this prediction is based on both theory (Rokeach, 1960; Suedfeld and Hagen, 1966; Berlyne, 1957) and earlier data (Wilson, 1965; and the exploratory study). No difference can be predicted in solution time between the true and the mixed items). For Transfer Effects H (Note: 4. An interaction between order of testing and true and false item content is predicted for solution times. Based on the Lefford (1946) study of errors, solution times should show a positive transfer from the false-then-true order, while the true-then-false order should show no significant transfer benefits. This hypothesis is based both on the exploratory study and an einstellung notion (Schulz, 1960). No effects of presentation order on number correct can be predicted, for no evidence indicated differential transfer on the exploratory study, and Lefford's interpretation does not appear to be soundly based--never replicated). A preliminary scaling of the premises used in the three test forms was done to insure that the subject population actually interpreted the material as it was designed. Thus a check on the experimenter's a priori judgment of the item premises was made. METHOD Subjects: Subjects were undergraduate psychology students at Michigan State University during the Winter and Spring 1967 terms; a total of 89 subjects participated in the scaling of the premises for truth- falsity, another 99 subjects participated in Experiment I (randomly assigned to conditions), and 24 subjects participated in Experiment II (randomly assigned to conditions). Material: The materials used in the premise scaling for truth-falsity consisted of two separate. scaling batteries (Form 3 has 117 items and Form 4 has 109 items). Subjects used IBM 10-choice answer sheets to record scale values for each statement, and the Michigan State University Test Scoring Office provided a complete frequency count for each statement directly“ from the IBM answer sheets. These materials may be found in Appendix A. In Experiment 1, three forms of a 27-item syllogism test were deve10ped. Similar to the technique of Wilkins (1928) and Lefford (1946) these test forms were designed to vary in the truth-falsity of their premises-—but the truth-falsity of these statements was also judged by a sample of subjects. Form I of the test was composed of all true premises; Form IV all false premises; and Form 1M items had one true premise and one false premise. No attempt was made to match all the items in these tests for item length, logical structure, or logical conclusions, though 15 of the 27 items are matched for structure and length and represent matched test forms differing in truth-falsity. These tests and their IBM answer sheets 12 13 may be found in Appendix B. In Experiment II, subjects were presented only 10 of the 15 matched items from Experiment I which were further selected to equate the three test forms for word length. Table 1 shows the comparison for word length for the three scales. Table 1 Mean Length and SD of True, False, and Mixed Reasoning Scales Used in Exp. 11 True Mixed False X words/ 41.1 41.1 40.9 item SD 6.04 7.58 11.00 These three forms of syllogistic reasoning tests are identical for syllogistic structure, and each form has three indeterminate items (# 4,8,10) and seven determinate items (all seven remaining). For presen- tation to the subjects, items were typed separately on standard 5 x 8 cards for use in the Johnson Serial Exposure Box. An example of an item from the ”True” premises scale is as follows: Varsity football is a very pOpular sport here at MSU. POpular sports have large crowds. l. Varsity football has large crowds here at MSU. 2. Varsity football is popular at Notre Dame. 3. MSU is a first-rate academic institution. 4. John Hannah likes football. 5 None of the above. The item from the ”False" premise scale matched with the above item for logical structure was as follows: 14 Physics is the most popular course of study here at MSU. Popular courses of study are very easy. 1. Physics is very easy here. 2. Elementary Education is easier than Physics. 3. Physics is very difficult here. 4. Physics students are very intelligent. 5 None of the above. The item from the "mixed" premise scale matched to the two proceeding items for logical structure was the following: Varsity football is a very popular sport here at MSU. All popular sports have very small crowds. 1. Varsity football here has large crowds. 2. Varsity football here has small crowds. 3. Varsity football here is not very popular. 4. Varsity football here provides much income for the university. 5. None of the above. Appendix C lists the items from the original 27 item tests which were selected for use in both the 15- item Experiment I analysis and the lO-item Experiment 11 analysis. Individual subject's records for solution time and choice, in Experiment II, were kept on a standard record sheet developed for this purpose, and this can be found in Appendix D. Apparatus: In Experiment 11, subjects were presented individual test items with the Johnson Serial Exposure Box. This apparatus allows measurement of premise encoding or reading time, conclusion selection time, and conclusion choice; and it has been described elsewhere in the literature (Beech, 1964 ; Johnson, 1962). Procedure: For the premise scaling task, subjects 15 were given the scaling batteries in a group testing situation of from 4-20 subjects; these were administered either by the experimenter or by regular teaching assistants during class time. The following instructions were given for this task: I "You are to read the following statements and rate or score them by marking on the accompanying IBM answer sheet a number from one to seven on the scale which you will be instructed to use. These statements may vary from one end of the scale to the other; your own judgment is the best answer. A statement neither one way or the other or indeterminate should receive a score of four. Here is the scale you are to use--the number one end of the rating scale is the completely true or valid end; the number seven end is the completely false or invalid end. You are to rate each state- ment for its validity or truth-falsity. Rate each statement on this scale, work as rapidly as possible, but make sure you do not lose your place on the answer sheet. The IBM answer sheets were then scored for a frequency count for each item, and each item received a truth falsity value based on a semi-interquartile range + . . . . . of -1. For example, the follow1ng distribution might have been obtained on an item: Scale Value: 1 2 3 4 5 6 7 Frequency 2% 8% 14% 40% 11% 15% 5% For this hypothetical item, a scale value of 4 would be assigned, for t 1 from 4 includes at least fifty percent of the distribution of ratings. This survey scaling of the item premises was used to verify the a priori development of the true and false premise syllogism tests, and to insure that items in Experiment I and 11 did 16 differ significantly and consistently on the truth—falsity variable between the three test forms. Experiment I was designed to compare tests with three types of item content in terms of number correct. All subjects were given the following instructions by the experimenter: You will be given a series of arguments or statements from.which conclusions may be drawn. You are to choose the correct logical conclusion for each argument or set of statements from the alternatives listed. If none of the conclusion statements is logically correct, you would select the ”none of the above” alternative. Do not make any markes in the test booklet please. Work rapidly, but do not rush carelessly through the material. Answers were recorded on IBM answer sheets, and tests were scored by the Michigan State University Test Scoring Office. In Experiment 11, the principle interest was the collection of data on total solution time for the three types of item content. Subjects were tested in individual testing sessions and randomly assigned to conditions; each subject worked two ten-item syllogism tests in one of the following orders: TF, FT, MT, or MF. This treatment design was used to provide two different comparisons: 1. analysis for differences between the three test forms for both number correct and solution time when items have been carefully selected for word length and logical structure, but differ in truth-falsity; and 2. analysis for relative transfer effects between the true-then—false and the false-then- true order of presentation. Eight subjects were used in each treatment order. 17 Originally it was decided to eliminate any subject with prior training in syllogistic reasoning, and one subject in the initial 24 was rejected when post- experimental questioning revealed that he had received training in syllogistic reasoning quite recently. Each subject in Experiment 11 was given the following instructions by the experimenter: Here is what we will do in this experiment; I will present items similar to the one on this sheet of paper (show subject sample item) with this apparatus (Show subject where item will appear in window). Sample item: All patriots are loyal. Some Frenchmen are not loyal. a. Some Frenchmen are not patriots. b. Degaulle is a French patriot. c Most Frenchmen are loyal. d. Brigitte Bardot is a very loyal "Frenchman." e. None of the above. You are to use the first two statements you are given as your background information; these will appear in the left hand window of the box when we begin. After you have read and under— stand the material on the left, flip the toggle switch (point to switch) to your right and carefully examine the statements you see there. If one of the given statements is a logical conclusion or deduction based on the background information you have been given, press the button on the box panel corresponding to that selection; if no alternative is correct, press button #5. If after examining the conclusion alternatives, you desire to reread the background information, you may do so by returning the toggle switch to your left. You are in complete command of how long you spend examining each part of the problems you will be working on. Do you have any questions? 18 The inter-item presentation interval was approximately 30 seconds, and the experimenter gave no signal that after the first 10 items had been completed a second test was being given. All tests were ‘administered in a standard order of presentation of items within tests, and the experimenter recorded premise reading time, conclusion selection time, and the alternative selected on the individual subject's record sheet. RESULTS The examination of the truth-falsity scaling of the premises used in the test forms to be compared in Experiment I and II indicated that the premises assumed by the experimenter to be false and those assumed to be true were so interpreted by the subjects. Using the mean scale value for the two premises in each item, EXEEX item in the true premise battery had a scale value from 1-3, while every item in the false premise battery had a scale value of from 5-7. Effect of Premise Content on Number Correct From Experiment I, the results with 27-item scales which differ in truth-falsity, but are not equated for logical structure, may be found in Table 2. Table 2 Mean Number Correct by Premise Content, Exp. 1, 27-Items Mixed True False E 13.94 17.06 19.12 SD 5.63 2.15 3.08 Table 3 summarizes the analysis of variance for any overall differences in the effects of the three types of tests. Table 3 ANOV for Total Score on 27-Item Tests Source S.S. ‘gf M.S. F. p Tests 449.2 2 224.6 14.27 .001 Error 1511.2 96 15.7 Total 1960.4 98 19 20 The differences between the means for the groups (shown in Table 2) were analyzed with ordered t tests, and each mean is significantly different from the next in the series (between M and T, t=2.94, p<£001; between T and F, t=3.12, p<..001).1 The analysis of the effects of premise content on number correct isconfounded in the 27-item scales by the effects of items unmatched for logical structure. A second analysis on the above data can be made on the 15 items in each of the scales which have been matched for logical structure. These 15 item subsets of the original scales may be compared for mean number correct with equated sets of items, and Table 4 summarizes this comparison. Table 4 Mean Number Correct by Premise Content, Exp. 1, 15-Items Mixed True False “i 7.88 8.88 10.39 SD 3.39 1.30 1.80 These results for matched scales were analyzed for overall differences, and Table 5 gives this analysis of variance. l. Duncan's Multiple Range Test corroborates these results--between M and T, pe C. 3 cl- (5‘ "\ I O U P!- '3 :3 9‘1. 1 ‘c. .4 J O 1...; w P) III I U Q 0 (‘\ L a l S C . 1 3 so :1- if p" 2'“ Ci .2 2° F'}.;:.‘eies is the tweet 1er‘.‘:i-.‘~.i‘ wearer: cf‘P 811121;. Thus 3.1;- 15.8120 1; L2. ‘13! (.1...‘ y‘.‘.’5 ,!1. 1 p 4!. P .p; 1? e. ’5 m‘ I fix 1.. 1» ' { p ‘L - . . '3’ -‘--. F,~". - ‘ ‘. 1,11%. -“.,§‘«1 : -‘ L- 3.]..1» \o 2-1..- '-' ‘u (JJ..).' .. he .- .1 :1 . (fl-I. - v0 J. . .t ..L\. J.) ‘A only .34; e vmxgn v- .‘a‘. -4...ero w a. a _ . . — K; a 3.. I J “I - ; . ~ 1 J ‘ L-If ‘e \1 V“ I fix {‘1 ‘ J. L 'r . : 0.; lr‘ .) b5. 'i f‘ o - —. ‘.. r u f r ‘ ‘ c a. ‘1. . a 1 f. -:-. ‘d w‘ e ) 6 s.""‘ Whig"! p.133, .' 1-; 1:." -3 3 ,3; (.s :7. 1' i! (I \ ‘ 1' , -- ‘ . .. - . ¢ , 1n} . 1 . ”J .121 c. C: 8.. ,: a 7 r v- q A O I l “' '4 1 a- ’ nr * s K r ‘ L) A“ t A 3 f“ i ‘. L b“ k' :- l‘ l c ‘ 3 — I " V) 'H 'l ‘ ll . - ’ 'v 1"." " Q ‘3 "‘ ‘ "‘ 5‘ . k‘ ’j An ‘0 .‘. 3., '1 1.1.6 .-‘;‘a.l--k - ~ v {I l ‘4. (.4 ‘- J I 3'- ‘. a! * ‘24 I '3 Q -1 \«~'$‘ 9*“er -‘r\ " L . ‘ l. '7 " 31' ’ .1 .0" ‘4" '1."- I’ll ’ L 0 L (’1‘ ‘1: ‘4 a ' ‘J l r - i) -‘~f - . ' 31“ ‘A l m 6. v ?. Sane o: ”it: 1".” 97-9sz ? g. '7 .~- ‘2. A12 Calucge students fi1* J-'uffit?. 113 "ali;t“3 m “M r‘ ”“530- s l , ,1 —: "a. all] 1“". :MJ 8 '7 “o " anther? :f‘ L: L -.3 '2 flick; :1: rug; in")? Q.O 7.; ‘17'51 (u' ‘1"! .3 ‘1‘?" 2.". 1“} .41 :5“ x :‘I 75]";“1‘gr'1'1'rh I,’I§:.'"-.l,i #31 £1? 7'}??? T3}; 3‘7? 5:1 L11. ..‘\‘.... .' _‘.ev-.‘.....; g)‘ d -.__.| .L'; J .'.,..a 2 Li -.J ,J. -' .-. .. .-...*...'».r.- 4.13.. .1. .. .. v r . \ ,.. as. .4 us. 4 r1 F" ' ‘1‘ ‘§“‘ - "a - " w ‘v w- i " - -.:rO Off). 90 ”-1111 {7.1 L: 3):.‘30 l a X Page 2 Remember that you are rating these statements on the following scales (completely (%-&-5) (completely) 15. Some idiots are in mental hospitalso 16o All servicemen may be sent to Vict Name 170 Popular courses of study are very easy. 18. Some doctors (MoDe) are surgeonso 190 All popular sports havs_very small crcwdso 200 Only astronauts participate in the actual flying of space vehicleso 21° Anything educational should be studiedo 220 All liquid is solido 25° All criminals are lawbreakerso 24a 511 metal is veto 250 All Democrats are polo playerso 269 No vegetables are limeso 27° All potatoes are vsgetablcso 28° No girls are maleo 290 No books are papero )0. No amateurs are athleteso 31° No weapons are atomic water purifierso 52. No snow is weto 55. All students are zombieso 54o No rectangles are squareso 55o A11 cats are animalso 560 No Wingless creature can flyo 57o A11 patriots are loyalo 58o All boxes are roundo CHECK TO MAKE SURE YOU HAVE JUST FINISHED ITEM 58 ON THE IBM ANSWER SHEET 1! Go on to the next page £1! Iago Q {3 Remember that you are rating these statements on the following scale: 1 - 2 - 5 a 4 - 5 - 6 - 7 (completely (% &-%) (completely) 59. Some wonderful and intelligent men should be given international recognitiano #0. Flying over the water requires wings and a powerful engineo 41. All revolutionary radicals should be arrestedo 42. Some solid things are very brittleo #50 Some vegetables are sickeningo #4. All newspapers are papero 45. All athletes are interested in physical fitnesso 46. Some wet things are colda 47o Sore dags are eaten #8. L11 narcotics are stimulantso #9; Some criminals are sicko 90a All male college students may be drafted. 51. All lines are greeno 520 All boys are males 550 All atomic water purifiers are nuclear deviceso 54. Some animals are uinglesso 55° Some Frenchmen are not loyal. 56..All felines are animalso 57o Anything round is a cylinders 580 Some liquid is not veto 59¢ Some Democrats are mono 60° All athletes are professionalo 615 Some zombies are scholarso 620 Some circles are rectangleso 630 Some partitions are floorso 64. All boxes are barrelso 650 All physical pain is a stimulanto CHECK TO MAKE SURE YOU HAVE JUST FINISHED ITEM 65 ON THE IBM ANSWER SHEET 1! l‘ 0.0".) ‘1 w‘ Remsmfer that you are rating these statements on the following scales ‘1 "9 2 a 5 a" I! e» 5 v 6 c, 7 (completely) (% & ,) (completely) 29 (c 66, All autobiographies are biographies0 670 No women are queenso 68° No queens are womeno 690 Some lies are interesting storieso 70° No females are cowso 71o Some elephants are gray animals 72° No roses have thornso 73. Some trout are fisho 740 All donkeys are man-eaterso 750 No violets are roseso 76; All assistant professors are faculty memberso 770 Nb chickens are birdso 780 No dcgs are animalso 79o Nb bussos are carrierso 80° All voters are femaleo 810 Ne snow is veto 820 All men are animalso 850 All Americans are human beingso 8kg All animals are humanso 859 All snowmen are abominableo 869 No partitions are wallso IOU SHOULD HAVE JUST COMPLETED IBM mama 52m? NUMBER 1 mass 86 IS THE LAST STATEMENT ON THE FIRST.KNSWER SHEET 9!!!!! iNOW GO ON TO IBM ANSWJR SHEET NUMBER 2, AND CONTINUE EXACTLY AS BEFORE !!!! 1. All barrels are cardboardo 'Q 2. All drugs are stimulantso CHECK TO MAKE SURE YOU'HAVE JUST COMPLETED ITEM 2 ON THE SECOND IBM ANSWER SHEET I! Then, Go on to the next pagell Remember that you are rating these statements on the following scales (completely (31-; 8o ') MB 3.: All biographies are bookso 4° Some secretaries are queenso 50 No lie is trueo 60 No elephants are catso I. All roses are thornyo 80 All cows are giraffeso 90 Ne gray animal is a mammalo 100 He fish are chicken-o 11. All mama-eaters have four legso 12° All department heads are faculty memberso- 15. All thorny plants my scratch small childreno 1110 Some roosters are ahiokenso 150 All animals are organisms. 160 All carriers are motor vehicleso 17.. Some women are not femaleo 18. All white objects are abominableo 19. Some human beings are blondso 20o All cstriehes are animalso 21. All wet things contain watero 22.8mm females are under 21 years of ageo 25. Some humans are bald-headedo Page 5 C) 7 (completely) THIS IS THE END OF THIS TASK - MAKE SURE YOUR LAST RESPONSE WAS ON ETEM NUMBER 23 ON THE SECOND IBM SHEET w THANK YOU VERY MUCH H! 1 run vs-IISOI .\'1 [CH IGA N STATE UNIVERSITY RR‘AE DATE STUDENT NO. SEX M F LAST FIRST MIDDLE GUJRSE NAME COURSE NO. * S T U D E N T N U M B E R SECTION INSTRUCTOR 7an ____|_ 2 3 “4., _5 6 7 8 9 ‘IQII'E OF TEST FORM _L.-- ;:_;; ..... :_‘_; _______ _,7-__ ~~~~~ :;;;: ;;:_': Li... :5 SURE YOUR MARKS ARE HEAVY AND BLACK 0 . _2 3 4 g e 7-- a, 4 ERASE COMPLETELY ANY ANSWER CHANGED o r 2 _3 4 _3 __4_ _3 a 9 'l I I I I N I I I I 0‘ II I I I & I 0' I II I 0’ I I I I q II II II (D I I ‘0 | I | .I 0 II I I I II N O I! I. N I I I I 0‘ II I! II I! II bI I l I I U II I I | II OI II .I II II N II II II II E II II II II II ID I | 5 I I 1.! II I I N I II I! u II b II II I I 0 II I I I I d5 | l l I V II II I I. (D I II ‘9 I I II II 5 I I II II I LN Q3 I l I I N I I I OI II I I. I: b I I II M II II |I II 0'9 II II II II II V :I II II 0 .I 'l cl II II CD I! II I I 5 I I II II II IT‘I:T.: 2 ;‘::: 3 f:::: 4 ::;;- 5 1.77:: 6 :‘.;:: 7 TT'72 8 ::_'I: 9 _-;:, l0 t;.;: 77 |':::: 2:1:77 3 :’:,’.: 4 ;'_;:: 5 :-::' 6 ":__ 7 4 ’ IO -_;:: """ ; 6 :r::: 7 :::;; 8 .‘TZI‘; :::::IO ::::: 86 I :f":: 2 : 1:: 3 ::f:: 4 37:: 5 :i::: U! l I I I I II II I APPENDIX B C I ‘ I“. .‘e 2" .’_ h I“ t - L . .‘ a,‘ V! I -- I. 44 «. I,» .. J I L1 a I I J :2 ..I I4- - .‘J a . ‘ r' a '4: .3“ - n ‘3‘ I ~ 4;. - - ..3 '.I «I: ~33 ..w -- r- - .3 .- (1.. I J ‘ was (_ "5 1 ’ 3‘ - "’. 7 .“ k' '“I’ 3 it: .345. L — I . ,I 2 J 2 a ’.\ {Iqae tyla asoims popu5ar music In plared on thy raaioo radio music is quite enter: a1 zin >>>v 0‘ m Ié‘l‘ .J-‘- }-:-l ’4 9‘ v Thu grplant :Lat ate Chicggo” 13 $13ko H Cu'rnnI; popuier magi? i.. 33 a? go n; a novel" phaser” a“ A; 39F mu; uh: I q C tertnlr Ag. 1. Ta; CHEy wzy 1; bsv :IIus- 'irg RAILC 1; ts I3v aggc Egg 13gpffigV e, F“:0 of tld a7_:es it A; 0? hast ;. r ,3 533 :{;i V, ”a fi‘L 32y land Erna 3Lr-'33ia. 1 VA a In '3 3.an3 (or 1 tortirvlt‘( a”, AQ.I338.,.E 15 a -333: 3‘ :3.; r ‘ 4: Jrfimna_Iv3 11 33., 3323+..ch T8. is E .4‘3FJ.I.:~* "3; 23,11 43;: c... A‘LI‘Itrjais .18 15 h: ”7.3.1 I. (<.‘-.' 3‘. :..r.;fi37331;'}0 d Aufiiray a is t w 13' a to :0 1V '; and (:“ a 901;.nert ad Home Cf tqa a11n3 4a L3OIJJ pz33rclx Iczg‘ ..4 ~ ;.Id;r :33t; ,. 1: 1(47 Sen: staiiea of V? I‘ir’ *Jr 3ru t?:viti, unimpazfiazh resonrnh: an Al- TEECFOICF“ E v 3,: to: .,3h tiqu bu Some rbi Deh31.3‘ 1: ant RE fimpu4futt as htmm. a bekavicrh ;. Th: 53‘.331C~':‘ r if cf 3 3n twhnvior Exagta Ihc izasrest of rCra FSWCJO‘oIiRfiB. d) Page 3: Chain: :‘R d; irir a . *“impn Cant rauearrh- e “owe C- t}: Ev311I 5. H5 90.. icl“.3 x? Z tn r 131' .n er'CG forainr 3;; dispatqws 2:32 C .0 :orEiI i 3.2n3 fornuurc an 13 p-11fi’: an c a 2‘359134_ b; Dicfizf3 C ar: i?;guc Ciy as: 3333Afind. an F0 iIu‘Rl? fiad;1r‘d far rt 3 3:3b3L (1.. LI). 51176-- 4672‘: Md “Eitzciavfi 2.; 33:31:13,214 13 3.1" “MI t3 1'” Cut'hl'fgi Y) 3,, I": 8 0? tin 113.333,, C- A?) ..C( tIralI 33am I3 are) S‘XE‘OEIQ‘ m3. RIII’nyJ «3 yang anI hgaitIlV 313p e ara ru E an wa a yiflga In in C.1ig§fl9:. 3u2:-] I30f67310n81 Zoo:h&11 gi‘gygzs ura vual paidg bu All iaotball playars waar C?‘i5r:i”3 0:135fl1fi3 wills p255-“52 a: A1.1 strong and haaiIEy fooihh l Myn?3 would like ta 13 prcfasai~ 5 d2 all faouball pla;lars a‘e mv 3 aha?» average In inIoIIipearuv 6- None of the aboveo To It has been found tha, all rtca; C 2 .. 2.3 3 -? 0-1.; (1? '— Anyihing which contEiqa Iran 83 Steel in quiia dureEivq b9 Bath iron and steal '1 ,f ' 11.3.. (I 3:.) Steel 9 rrgai. me rather -4: V1235 ’3; 1133170 do Iron is the basa f2; firvy wsiao QGIH naufit-C §~ 'I‘xx 73”“ (:1 "X; (N \J C70 b~3 \ y . I. .IIIII .4.?IILI.I‘ .u .o.‘oA-I ,I‘avn‘t‘ .fi.” 14‘5ng L . '. II 1' I‘ll! .ui ‘11 . ”.141 ”VIII, “p “g” _. ._ 1 .m I’In SP . flow—mm 43.33130, 3 .3 hag—<11}. . .m Swwrfi- N54; 1 \ I. . ....I'II ....I .«34. .,..Q’F4..:I'.Lv.rl arr/m...” AVHVJRH Inna, m I LI... MID . m m E . M. m w I M W .. m in. w .2. w .mI “3 IN AN Q .3 . .6. a 3. ._ _ .w E _m A .r (~14); NI m»? mm...